Dataframe regression with tolerances for numeric and not just float values

[HikingDev]

Currently the check performed on numeric values checks if the obtained column dtype is of subtype np.inexact.
This excludes tolerances for Integer values. Is there any specific reason to exclude Integers?

E.g. the type of a column is Int64 and the unit (not shown of course) is nanoseconds, the tolerance check is excluded.

Rather than dataframe_regression#L127

np.issubdtype(obtained_column.values.dtype, np.number)

[nicoddemus]

cc @tadeu @tarcisiofischer

[tarcisiofischer]

AFAIR, it was because we only used for floats... Feel free to add it, though. Sounds like a good idea to me

[seanv507]

rather than doing this, what about supporting pandas assert_frame_equal: https://pandas.pydata.org/docs/reference/api/pandas.testing.assert_frame_equal.html ?